AI Paper Atlas

Research map / Large language models / Retrieval and knowledge

Query-Efficient Domain Knowledge Stealing Against Large Language Models

Zhengao Li, Xiaopeng Yuan, Bolin Shen, Kien Le, Haohan Wang, Xugui Zhou, Shangqian Gao, Yushun Dong

AAAI 2026 · Large language models › Retrieval and knowledge · approach: Knowledge-augmented question answering · 1 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add chain-of-questioning (CoQ) (ablation)query-efficient stealing framework · MedQAaccuracy: 59.1 -> 62.0Improved
add chain-of-questioning (CoQ) (ablation)query-efficient stealing framework · FOMCaccuracy: 52.2 -> 54.0Improved
add feedback-guided scheduling (FB) (ablation)query-efficient stealing framework · MedQAaccuracy: 58.9 -> 62.0Improved
add feedback-guided scheduling (FB) (ablation)query-efficient stealing framework · FOMCaccuracy: 53.9 -> 54.0Improved
add cluster-guided exploration (ablation)query-efficient stealing framework · MedQAaccuracy: 60.8 -> 62.0Improved
add cluster-guided exploration (ablation)query-efficient stealing framework · FOMCaccuracy: 53.4 -> 54.0Improved
use structured exploration instead of EvoKDLLaMA-3.1 8B surrogate · FOMCaccuracy: 43.5 -> 54.0Improved
use structured exploration instead of Model LeechingLLaMA-3.1 8B surrogate · ChemProtaccuracy: 44.7 -> 43.1Hurt
use structured exploration instead of Model LeechingLLaMA-3.1 8B surrogate · PubMedQAaccuracy: 74.3 -> 74.9Improved
use structured exploration instead of Model LeechingGemma-2 9B surrogate · FPBaccuracy: 60.1 -> 68.5Improved
use structured exploration instead of EvoKDGemma-2 9B surrogate · HeadLineaccuracy: 78.5 -> 80.4Improved
train surrogate on 500 stolen QA pairs vs untuned baseLLaMA-3.1 8B · ChemProtaccuracy: 24.7 -> 43.1Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy